CARD: Databasing Real-Time Collaborative Reconstructions of 3-D Neuronal Morphologies

Reconstructing complex three-dimensional (3-D) biological objects (e.g., neurons) from bioimaging data is a key technique in bioimage informatics, with a wide range of applications in biology. Traditional data management systems for neuron reconstruction typically support only static data exports after proofreading, while recent collaborative platforms that enable real-time analysis and proofreading focus primarily on electron microscopy data. Neither adequately addresses the need for dynamic, large-scale collaboration on light-microscopy neuronal morphology. Here we present the Collaborative Augmented Reconstruction Database (CARD), a distributed infrastructure designed for real-time, multi-user collaborative proofreading and annotation of light-microscopy neuronal data across heterogeneous devices. A core technical innovation of CARD is ConMemMap, a contiguous-memory hash map that replaces conventional node-based hash tables with a compact, cache-friendly layout, substantially improving insertion, querying, and deletion throughput while reducing memory consumption. Built upon ConMemMap, CARD integrates real-time quality control for detecting reconstruction errors such as loops and missing branches, a version control module combining user-created snapshots with incremental operation records, and comprehensive metadata management spanning species, brain regions, annotators, devices, and timestamps. Benchmark evaluations demonstrate that the system maintains low latency under high concurrency, supporting efficient collaboration among dozens of simultaneous users on reconstructions containing millions of nodes.